CN115014299A - Flood peak early warning method based on Internet of things and big data - Google Patents

Flood peak early warning method based on Internet of things and big data Download PDF

Info

Publication number
CN115014299A
CN115014299A CN202210953183.3A CN202210953183A CN115014299A CN 115014299 A CN115014299 A CN 115014299A CN 202210953183 A CN202210953183 A CN 202210953183A CN 115014299 A CN115014299 A CN 115014299A
Authority
CN
China
Prior art keywords
early warning
water level
river reach
level
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202210953183.3A
Other languages
Chinese (zh)
Other versions
CN115014299B (en
Inventor
杨牧
吴西贵
杨江骅
母利
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Mountain Technology Co ltd
Original Assignee
Mountain Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Mountain Technology Co ltd filed Critical Mountain Technology Co ltd
Priority to CN202210953183.3A priority Critical patent/CN115014299B/en
Publication of CN115014299A publication Critical patent/CN115014299A/en
Application granted granted Critical
Publication of CN115014299B publication Critical patent/CN115014299B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C13/00—Surveying specially adapted to open water, e.g. sea, lake, river or canal
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C13/00—Surveying specially adapted to open water, e.g. sea, lake, river or canal
    • G01C13/002—Measuring the movement of open water
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
    • Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A50/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE in human health protection, e.g. against extreme weather

Abstract

The invention relates to the technical field of hydrologic monitoring data processing, in particular to a flood peak early warning method based on the Internet of things and big data. According to the method, three constraints are additionally constructed by a method for extracting a data relation, the first constraint is the possibility that various early warning levels are generated under current real-time data constructed after the deviation degree of the early warning levels of all water level meters of a river reach in historical data and the whole early warning level of the river reach is considered, the second constraint is the possibility that the early warning occurs in each month of the river reach according to the number of early warning times in each month in the historical data, the third constraint is the rainfall difference between other areas and the river reach to be early warned after the rainfall condition is determined to be similar to the rainfall condition of the current river reach to be early warned, the convergence speed of the neural network is accelerated by the three constraints, so that the neural network training can be completed by using less hydrological data, and the problem of insufficient early warning accuracy when the neural network is used for carrying out flood peak early warning is solved.

Description

Flood peak early warning method based on Internet of things and big data
Technical Field
The invention relates to the technical field of hydrologic monitoring data processing, in particular to a flood peak early warning method based on the Internet of things and big data.
Background
The flood disaster not only can cause great loss to agriculture, but also can cause serious industrial and life and property loss, so how to accurately early warn the flood disaster to prevent or reduce the loss caused by the flood disaster is an important problem to be solved in the flood disaster treatment.
The current flood disaster early warning method is various, the most common method comprises early warning by using a neural network, however, the neural network needs to be trained by using a large number of training samples, and the hydrological monitoring data has longer sampling time interval due to the consideration of effectiveness, so that the finally obtained total data volume is smaller, the neural network often cannot obtain enough training samples to complete sufficient training, and the early warning accuracy of the current adopted neural network to flood disasters is lower.
That is, the existing method for performing flood disaster early warning by adopting a neural network has the problem of low early warning accuracy.
Disclosure of Invention
The invention provides a flood peak early warning method based on the Internet of things and big data, which is used for solving the problem that the prior art cannot accurately early warn the flood peak by utilizing a neural network, and adopts the following technical scheme:
the invention discloses a flood peak early warning method based on the Internet of things and big data, which comprises the following steps of:
acquiring the water levels of all water level meters of the current river reach, determining the distance between the water level of each water level meter and a warning line of each level, and simultaneously acquiring the rainfall and the data recording time of the current river reach;
determining the correlation degree between the early warning level of each water level gauge in the current river reach and the whole early warning level of the current river reach, then determining the possibility of various early warning levels under current real-time data by combining historical data, and taking the possibility as a first constraint quantity;
determining the possibility of early warning in each month according to historical data, and taking the possibility as a second constraint quantity;
respectively establishing daily average rainfall capacity and early warning grade relation curves of the current river reach and other areas, calculating the similarity between the corresponding relation curves of the current river reach and other areas, determining other areas similar to the early warning condition of the current river reach and taking the other areas as reliable areas, calculating the difference value between the historical daily average rainfall capacity corresponding to the reliable areas when the early warning grades are generated and the current river reach rainfall capacity, and taking the difference value as a third constraint quantity;
and constructing a loss function of the neural network according to the first constraint quantity, the second constraint quantity and the third constraint quantity, finishing the training of the neural network according to the constructed loss function, inputting hydrologic data of the current river reach acquired in real time into the trained neural network to obtain a corresponding early warning grade, and finishing the flood peak early warning of the current river reach.
The invention has the beneficial effects that:
considering that the hydrologic data volume is generally small and the training of the neural network used for flood peak early warning cannot be well completed, when the loss function of the neural network used for flood peak early warning is constructed, the method additionally considers the constraints of three aspects to improve the conventional loss function so as to further constrain the neural network to accelerate the convergence speed of the neural network and improve the training effect; the first aspect of the constraint specifically includes that after the deviation degree of the early warning level of each water level gauge of the river reach in the historical data and the whole early warning level of the river reach is considered, the possibility that a certain early warning level occurs in the constructed current real-time data is considered, the second aspect of the constraint is that the early warning probability of the river reach in each month is obtained by combining the number of times of early warning occurrence of each month in the historical data, and the third aspect of the constraint is that after other areas with rainfall conditions similar to those of the current river reach to be early warned are determined, the calculated rainfall difference between the other areas and the river reach to be early warned is obtained. By the aid of the constructed constraints of the three aspects, the method can accelerate convergence speed of the neural network used for flood peak early warning in a training process and improve training effect, accurate flood peak early warning can be completed by means of the neural network under the condition that hydrological data volume is small, and early warning accuracy in the process of adopting the neural network to early warn flood disasters is improved.
Further, the determination process of the first constraint quantity is as follows:
calculating the importance degree of the water level data obtained on each water level meter:
Figure 413052DEST_PATH_IMAGE002
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE003
SW is the water level of the water level gauge, SJ is the probability of the category of the current water level after the water levels of all the water level gauges are clustered,
Figure 408821DEST_PATH_IMAGE004
represents the maximum of all the water levels,
Figure DEST_PATH_IMAGE005
representing the maximum value of the occurrence probability of various water levels;
determining the deviation degree of the corresponding water level early warning level of each water level gauge and the whole water level early warning level of the river reach:
Figure DEST_PATH_IMAGE007
Figure 437826DEST_PATH_IMAGE008
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE009
the deviation degree between the water level early warning level corresponding to the ith water level meter and the whole early warning level of the river reach is shown, T is the total number of the water level early warning levels obtained by the ith water level meter in a set time period,
Figure 938427DEST_PATH_IMAGE010
indicating the corresponding water level of the ith water level meterThe number of times that the alarm level is inconsistent with the whole early warning level of the river reach,
Figure DEST_PATH_IMAGE011
indicating the deviation of the early warning level of the water level corresponding to the ith water level meter and the integral early warning level of the river reach,
Figure 861122DEST_PATH_IMAGE012
showing the correlation between the j-th water level early warning level on the ith water level meter and the whole early warning level of the river reach,
Figure DEST_PATH_IMAGE013
indicating the j-th water level early warning grade on the ith water level meter,
Figure 558950DEST_PATH_IMAGE014
representing the integral early warning level of the river reach of the drainage basin;
determining the possibility of various early warning levels under the current real-time data by combining historical data:
Figure 763667DEST_PATH_IMAGE016
Figure 136136DEST_PATH_IMAGE018
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE019
indicating the likelihood of an early warning level of p,
Figure 680381DEST_PATH_IMAGE020
the distance between the water level corresponding to the estimated water level line and the warning line is obtained by the sum of the water level at the ith water level meter and the rainfall which are acquired in real time,
Figure DEST_PATH_IMAGE021
corresponding to the early warning level on the ith water level gauge when the early warning level of the whole river basin in the historical data is pThe distance between the water level and the warning line,
Figure 782066DEST_PATH_IMAGE022
the total years of the history data used, Q is the number of water level meters,
Figure DEST_PATH_IMAGE023
in order to be the age-related weight,
Figure 880604DEST_PATH_IMAGE024
is a threshold value for the trustworthiness of the data,
Figure DEST_PATH_IMAGE025
is the first in the history data
Figure 457429DEST_PATH_IMAGE026
Early warning level of year.
Further, the determination process of the second constraint quantity is as follows:
the probability of early warning occurring in each month is:
Figure 652918DEST_PATH_IMAGE028
Figure 191084DEST_PATH_IMAGE018
wherein, the first and the second end of the pipe are connected with each other,
Figure DEST_PATH_IMAGE029
indicating the likelihood of a pre-alarm occurring at month m, Y is the total years of historical data employed,
Figure 42497DEST_PATH_IMAGE030
indicating the number of occurrences of the pre-warning at the mth month of the u-th year,
Figure DEST_PATH_IMAGE031
indicating the total number of occurrences of the pre-alarm in the u-th year,
Figure 428872DEST_PATH_IMAGE023
in order to be the age-related weight,
Figure 744447DEST_PATH_IMAGE024
in order to be a threshold value for the confidence of the data,
Figure 515832DEST_PATH_IMAGE025
is the first in the history data
Figure 182436DEST_PATH_IMAGE026
Early warning level of year.
Further, the determination process of the third constraint quantity is as follows:
establishing a relation curve between daily average rainfall and early warning level of the current river reach and other areas, and calculating a dynamic time normalization distance between the corresponding relation curves of the current river reach and other areas:
Figure 100002_DEST_PATH_IMAGE033
wherein the content of the first and second substances,
Figure 136617DEST_PATH_IMAGE034
the dynamic time integration distance between the average rainfall-early warning grade curve of the current river reach in the past year and the average rainfall-early warning grade curve of other areas in the past year is shown, R is the total number of element points on the average rainfall-early warning grade curve in the past year,
Figure DEST_PATH_IMAGE035
is the minimum cost between corresponding points of the two curves;
determining the reliability of other areas same as the rainfall condition of the current river reach:
Figure 100002_DEST_PATH_IMAGE037
wherein E is the reliability of the rainfall condition of the current river reach being the same as that of other areas, and Y is the adopted rainfallThe total number of years of the historical data,
Figure 435528DEST_PATH_IMAGE038
the dynamic time normalization distance between the daily average rainfall-early warning grade curve of the current river reach corresponding to the u-th year in history and the daily average rainfall-early warning grade curve of other areas is calculated;
and taking other corresponding areas when the reliability is greater than the reliability threshold value as reliable areas, and calculating the difference between the rainfall amount corresponding to each early warning level of the reliable areas and the rainfall amount acquired in real time by the current river reach:
Figure 879279DEST_PATH_IMAGE040
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE041
the difference value between the historical daily average rainfall and the current river reach rainfall collected in real time when the early warning grade of each reliable area is p,
Figure 672660DEST_PATH_IMAGE042
for the number of reliable areas to be used,
Figure DEST_PATH_IMAGE043
is as follows
Figure 931997DEST_PATH_IMAGE044
The reliability of the individual reliable regions is,
Figure DEST_PATH_IMAGE045
is as follows
Figure 628689DEST_PATH_IMAGE044
The historical daily average rainfall when the early warning level of each reliable area is p,
Figure 508920DEST_PATH_IMAGE046
the rainfall of the current river reach is collected in real time.
Further, the loss function is:
Figure 383073DEST_PATH_IMAGE048
Figure DEST_PATH_IMAGE049
Figure 100002_DEST_PATH_IMAGE051
Figure DEST_PATH_IMAGE053
wherein the content of the first and second substances,
Figure 518869DEST_PATH_IMAGE054
in order to be a function of the loss of the neural network,
Figure DEST_PATH_IMAGE055
is a cross entropy loss function, namely a loss function formed by the early warning level output by the neural network and the early warning level of the label corresponding to the input data,
Figure 961745DEST_PATH_IMAGE056
the probability that the integral early warning grade of the river basin is p is obtained when historical data is used for artificial analysis when the output result of the neural network is the early warning grade p,
Figure DEST_PATH_IMAGE057
indicating the possibility of the early warning result output by the neural network occurring in the current month and the possibility of the no early warning result output by the neural network occurring in the current month,
Figure 183517DEST_PATH_IMAGE058
the early warning grade of each reliable area is p, the corresponding historical daily average rainfall and the current river reach rainfall acquired in real timeThe value of the difference between the two values,
Figure DEST_PATH_IMAGE059
early warning the flood peak for the possibility of early warning in the current month,
Figure 656218DEST_PATH_IMAGE060
the times of early warning the flood peak in the current month with the early warning level more than 1,
Figure DEST_PATH_IMAGE061
the total days of the month are pre-warned for the flood peak,
Figure 720339DEST_PATH_IMAGE062
the days that the month has elapsed are warned of the flood peak.
Drawings
Fig. 1 is a flowchart of the flood peak early warning method based on the internet of things and big data.
Detailed Description
The specific scenes aimed by the invention are as follows:
when the flood peak early warning is completed by using the neural network, a large amount of data is often needed during training due to numerous neural network parameters, and the hydrological data in a certain area is acquired even once every hour every day, but the total amount of finally acquired data may be insufficient for the neural network, so that the situation that the flood peak early warning cannot be accurately completed due to low output accuracy of the neural network due to data lack, namely insufficient training samples, may occur during training of the neural network.
The flood peak early warning method based on the internet of things and big data is described in detail below with reference to the accompanying drawings and embodiments.
The method comprises the following steps:
the embodiment of the flood peak early warning method based on the internet of things and big data has the overall flow as shown in fig. 1, and the specific process is as follows:
1. and acquiring the water levels of all water level meters in the river reach of the drainage basin, the distance between each water level and the warning line of each level of water level, data recording time and rainfall of the whole river reach of the drainage basin.
Under general conditions, in order to guarantee the accuracy of the flood peak early warning, the water levels of all the water level meters in the river reach of the river reach are collected to finish the flood peak early warning, and the river reach of the river reach includes Q water level meters in the embodiment.
After the water level SW of each water level meter in the river reach of the drainage basin is obtained, the distance WX of the water level at each water level meter from warning lines of different levels can be correspondingly obtained, meanwhile, in the data of flood peak early warning, the early warning level YJ sent by the river reach of the drainage basin can be determined, meanwhile, the rainfall JY of the whole river reach of the drainage basin can be obtained, and the year, month, day and time Y-M-D of the water level meter for recording the water level can be obtained.
In this embodiment, the distance between the water level and the warning lines of different levels is divided into 5 types, that is, the value of the distance WX is 5 types, specifically:
normal water level, WX = 1;
near watch water level, WX = 2;
beyond the alarm water level, WX = 3;
exceeding and approaching the guaranteed water level, WX = 4;
past the historical maximum, WX = 5.
The early warning levels at the water levels correspond to the value of the distance WX between the water level SW and the warning lines of different levels, that is, the value of the early warning level YJ is also 5, and corresponds to the value of WX: no warning YJ =1, blue warning YJ =2, yellow warning YJ =3, orange warning YJ =4, red warning YJ = 5.
And recording data of the river reach of the drainage basin as G { H, Y-M-D, JY } when early warning occurs, wherein H is a two-dimensional array with the size of Q x 2, the first column of data in the two-dimensional array is the water level of each water level meter in the Q water level meters, and the second column of data in the two-dimensional array is the distance WX between the water level of the Q water level meters and warning lines with different levels.
2. And determining the correlation degree between the early warning level of each water level gauge in the river reach and the whole early warning level of the river reach, and then determining the possibility of generating a certain early warning level under the current real-time data by combining historical data.
Generally, in data processing and analysis, outlier data which obviously deviates from an average value is removed or ignored to a certain extent, but for the water level in the hydrological data information, the more the data deviates from the average value and has a smaller data amount, the more the flood peak disaster can be reflected in advance, so in the flood peak early warning, early warning needs to be performed according to the data which deviates from the average value, in the embodiment, the hydrological data which has a higher concern about the deviation of the water level from a normal value and has a smaller occurrence frequency is selected, and the importance degree of the water level data obtained on each water level meter is calculated according to the selected data:
Figure 453940DEST_PATH_IMAGE002
wherein the content of the first and second substances,
Figure 174509DEST_PATH_IMAGE003
SW is the water level of the water level meter, SJ is the probability of the category to which the current water level belongs after the water levels of all the water level meters are clustered,
Figure 524719DEST_PATH_IMAGE004
represents the maximum of all the water levels,
Figure 490401DEST_PATH_IMAGE005
and represents the maximum value of the appearance probabilities of various water levels.
The method for clustering the water levels of all the water level meters can use any one of the prior art, and the preferred clustering method in the embodiment is DBSCAN.
And determining the deviation degree of each water level meter compared with the integral early warning level of the river reach by combining historical data based on the obtained importance degree of the water level data on each water level meter.
According to the definition of the water level grade and the early warning grade, the water level grade corresponds to the early warning grade in a one-to-one correspondence mode. When early warning is carried out, one river reach is provided with a plurality of water level observation stations, and the water level data of a plurality of water level meters are integrated, so that the early warning level of the whole river reach is not corresponding to the water level of each water level meter, and the situation that the early warning level corresponding to the water level of a single water level meter exceeds the flood peak early warning level of the river reach or the early warning level corresponding to some water level meters is lower than the whole early warning level of the river reach can be realized. Therefore, when the flood peak early warning is performed on the river reach of the drainage basin, the deviation degree of the early warning level of the corresponding water level of each water level meter and the early warning level of the whole water level of the river reach of the drainage basin needs to be determined:
Figure DEST_PATH_IMAGE063
Figure 48814DEST_PATH_IMAGE008
wherein the content of the first and second substances,
Figure 910591DEST_PATH_IMAGE009
calculating the deviation degree between the early warning level of the water level corresponding to the ith water level meter and the whole early warning level of the river reach
Figure 748097DEST_PATH_IMAGE064
The larger the water level early warning level is, the lower the correlation degree between the water level early warning level of the ith water level meter and the whole early warning level of the river reach is, T is the total number of the water level early warning levels obtained by the ith water level meter in a set time period,
Figure 986311DEST_PATH_IMAGE010
the number of times that the early warning level of the water level corresponding to the ith water level gauge is inconsistent with the integral early warning level of the river reach of the drainage basin is represented,
Figure 724198DEST_PATH_IMAGE011
the deviation between the water level early warning level corresponding to the ith water level meter and the integral early warning level of the river reach is shown, if the deviation is more than 0, the jth water level early warning level of the ith water level meter is higher than the integral early warning level of the river reach, and if the deviation is less than 0, the jth water level early warning level of the ith water level meter is higher than the integral early warning level of the river reach, and the likeThe level is lower than the upper level of the power supply,
Figure 756876DEST_PATH_IMAGE012
showing the correlation between the jth water level early warning level on the ith water level gauge and the integral early warning level of the river reach of the drainage basin
Figure 81678DEST_PATH_IMAGE012
The more times of value 0, the closer the water level early warning level corresponding to the ith water level gauge is to the whole early warning level of the river reach,
Figure 619189DEST_PATH_IMAGE013
indicating the j-th water level early warning level on the ith water level meter,
Figure 447468DEST_PATH_IMAGE014
and the integral early warning level of the river reach of the river basin is represented.
And determining the possibility that the current real-time water level data correspondingly generates various early warning levels by combining historical data based on the deviation degree of the water level early warning levels of the water level gauges and the whole water level early warning level of the river reach.
Figure DEST_PATH_IMAGE065
Figure 854309DEST_PATH_IMAGE066
Wherein, the first and the second end of the pipe are connected with each other,
Figure 633784DEST_PATH_IMAGE019
indicating the likelihood of an early warning level of p,
Figure 479381DEST_PATH_IMAGE020
the distance between the water level corresponding to the estimated water level line and the warning line is obtained by the sum of the water level at the ith water level meter and the rainfall which are acquired in real time,
Figure 427745DEST_PATH_IMAGE021
the distance between the water level corresponding to the early warning level on the ith water level meter and the warning line when the early warning level of the whole river basin in the historical data is p,
Figure 569270DEST_PATH_IMAGE022
the total years of the history data used, Q is the number of water level gauges,
Figure 337506DEST_PATH_IMAGE023
in order to be the age-related weight,
Figure 986793DEST_PATH_IMAGE024
for the credible threshold of data, the invention prefers
Figure 524085DEST_PATH_IMAGE024
=15, i.e. data within 15 years of history are selected as credible data,
Figure 99160DEST_PATH_IMAGE025
is the first in the history data
Figure 354692DEST_PATH_IMAGE026
Early warning level of year.
3. And determining the possibility of early warning in each month according to the historical data.
As the flood period rules of river reach are mostly related to seasons, the possibility of early warning in each month is determined by the residual historical data with early warning after the data part without early warning is removed from the historical data:
Figure DEST_PATH_IMAGE067
Figure 733238DEST_PATH_IMAGE018
wherein the content of the first and second substances,
Figure 390615DEST_PATH_IMAGE029
indicating the possibility of early warning at month m, Y being the total years of the history data used, and a larger value indicating a greater distance from the current time,
Figure 638057DEST_PATH_IMAGE030
representing the number of pre-warnings appearing at month m of year u,
Figure 646464DEST_PATH_IMAGE031
indicating the total number of occurrences of the pre-alarm in the u-th year,
Figure 870510DEST_PATH_IMAGE023
in order to be the age-related weight,
Figure 382394DEST_PATH_IMAGE024
for the credible threshold of data, the invention prefers
Figure 800737DEST_PATH_IMAGE024
=15，
Figure 266747DEST_PATH_IMAGE025
Is the first in the history data
Figure 795948DEST_PATH_IMAGE026
Early warning level of year.
4. Determining other areas similar to the early warning condition of the current river reach according to the relation curve of the daily average rainfall and the early warning levels, and calculating the difference value between the historical daily average rainfall corresponding to the other areas when the early warning levels are generated and the rainfall of the current river reach, so as to determine the early warning level which is most likely to occur in the current river reach.
In some places with similar geographic positions, like adjacent river reach in a river, the height of the dam guard lines is similar, the weather conditions are similar, and the evaluation standards of various early warning levels are similar, so that the early warning conditions of the current river reach to be early warned can be determined by referring to the early warning conditions in the places.
In this embodiment, the similarity of the early warning conditions in the two regions is determined by calculating the similarity between the average rainfall-early warning level curves in the two places in the past year, specifically, the similarity between the average rainfall-early warning level curves in the two places in the past year is calculated by DTW, so as to determine the remote early warning data with reference value.
Figure 162338DEST_PATH_IMAGE033
Wherein the content of the first and second substances,
Figure 486004DEST_PATH_IMAGE034
the dynamic time integration distance between the average rainfall-early warning grade curve of the current river reach in the past year and the average rainfall-early warning grade curve of other areas in the past year is shown, R is the total number of element points on the average rainfall-early warning grade curve in the past year,
Figure 701959DEST_PATH_IMAGE035
the minimum cost between the corresponding points of the two curves. This embodiment is preferred
Figure 769272DEST_PATH_IMAGE034
The window size at which the calculation is performed is 40.
And determining the reliability of other areas, which is the same as the rainfall condition of the current river reach, by calculating the DTW value of each year of data in the previous year of data.
Figure 255748DEST_PATH_IMAGE037
Wherein E is the reliability of the rainfall condition of the current river reach same as that of other areas, Y is the total year of the adopted historical data,
Figure 511499DEST_PATH_IMAGE038
and (4) the dynamic time integration distance between the daily average rainfall-early warning grade curve of the current river reach corresponding to the u-th year in history and the daily average rainfall-early warning grade curve of other areas is obtained.
In this embodiment, it is preferable that the other region corresponding to the reliability E >0.1 is used as a reliable region, and data of the reliable region is used as referenceable data, and of course, in other embodiments, the reliability threshold may be set to other values; and then calculating the difference between the rainfall corresponding to each early warning level of the reliable area and the rainfall acquired in real time by the current river reach:
Figure 716216DEST_PATH_IMAGE040
wherein the content of the first and second substances,
Figure 321640DEST_PATH_IMAGE041
the difference value between the historical daily average rainfall and the current river reach rainfall collected in real time when the early warning grade of each reliable area is p,
Figure 161158DEST_PATH_IMAGE042
for the number of reliable areas to be used,
Figure 92205DEST_PATH_IMAGE043
is a first
Figure 518638DEST_PATH_IMAGE044
The reliability of the individual reliable regions is,
Figure 927754DEST_PATH_IMAGE045
is as follows
Figure 624708DEST_PATH_IMAGE044
The historical daily average rainfall when the early warning level of each reliable area is p,
Figure 726656DEST_PATH_IMAGE046
the rainfall of the current river reach collected in real time is obtained by counting the rainfall of the current river reach to be pre-warned on the day, the residual time of the current day and the forecasted rainfall in unit time.
Difference value
Figure 374806DEST_PATH_IMAGE041
The smaller the water situation is, the closer the water situation of the current river reach is to the p-level early warning level, that is, the more likely the current river reach is to generate the p-level early warning.
5. And constructing a loss function of the neural network according to the probability of a certain early warning level occurring under the obtained current real-time data, the probability of early warning occurring in each month determined by the historical data, and the difference between the historical daily average rainfall and the current river reach rainfall corresponding to other areas when each early warning level occurs, so as to complete training of the neural network and carry out flood peak early warning under the current water condition.
In the embodiment, a fully-connected neural network is used for flood peak early warning, data input by the network are two-dimensional vectors D { H, Y-M-D }, wherein H is a two-dimensional array with the size of Q x 2, the first column of data in the two-dimensional array is the water level of each water level meter in the Q water level meters, the second column of data in the two-dimensional array is the distance WX between the water level of the Q water level meters and warning lines with different levels, Y-M-D is the time when the water level meters collect water level information, and a training label is an early warning level YJ.
Because the data volume of the hydrological data is limited, and a sufficient training sample cannot be easily formed for the neural network, in the embodiment, the difference between the historical daily average rainfall and the current river reach rainfall corresponding to other areas when the early warning levels are generated is used as the training constraint of the neural network by taking the probability of generating a certain early warning level under the current real-time data obtained before, the probability of generating early warning in each month determined by the historical data, and the difference between the historical daily average rainfall and the current river reach rainfall corresponding to other areas when the early warning levels are generated, so as to construct the neural network loss function:
Figure 351727DEST_PATH_IMAGE068
Figure 667302DEST_PATH_IMAGE049
Figure DEST_PATH_IMAGE069
Figure 842282DEST_PATH_IMAGE053
wherein, the first and the second end of the pipe are connected with each other,
Figure 243307DEST_PATH_IMAGE054
in order to be a function of the loss of the neural network,
Figure 994226DEST_PATH_IMAGE055
is a cross entropy loss function, namely a loss function formed by the early warning level output by the neural network and the early warning level of the label corresponding to the input data,
Figure 429886DEST_PATH_IMAGE056
the probability that the overall early warning level of the river basin is p is obtained when historical data is used for artificial analysis when the output result of the neural network is the early warning level p is shown, the smaller the probability is, the better the output effect of the neural network is represented,
Figure 106593DEST_PATH_IMAGE057
indicating the possibility of the early warning result output by the neural network occurring in the current month and the possibility of the no early warning result output by the neural network occurring in the current month,
Figure 260494DEST_PATH_IMAGE058
the difference value between the historical daily average rainfall and the current river reach rainfall collected in real time when the early warning grade of each reliable area is p,
Figure 815103DEST_PATH_IMAGE059
early warning the flood peak for the possibility of early warning in the current month,
Figure 75576DEST_PATH_IMAGE060
the number of times that the early warning level of the flood peak in the current month is more than 1,
Figure 955808DEST_PATH_IMAGE061
the total days of the month are pre-warned for the flood peak,
Figure 65846DEST_PATH_IMAGE062
the days that the month has elapsed are warned of the flood peak.
And updating parameters of the neural network by using a random gradient descent algorithm according to the constructed loss function, finally obtaining a trained neural network, inputting real-time hydrological data of the river reach into the trained neural network, and obtaining a corresponding early warning grade, wherein the obtained early warning grade is the flood peak early warning grade under the current water situation.
According to the method, the relation among the data is extracted and is used as the neural network loss function, the improvement of the neural network loss function on the flood peak early warning is completed, the convergence speed in the neural network training process is increased, the problem that the convergence speed of the neural network training is low or even the convergence is not caused due to the fact that the hydrologic data volume is too small is solved, and therefore the accuracy of the neural network on the flood peak early warning is improved.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the embodiments of the present application, and they should be construed as being included in the present application.

Claims (5)

1. A flood peak early warning method based on the Internet of things and big data is characterized by comprising the following steps:
acquiring the water levels of all water level meters of the current river reach, determining the distance between the water level of each water level meter and each grade warning line, and acquiring the rainfall capacity and the data recording time of the current river reach;
determining the correlation degree between the early warning level of each water level gauge in the current river reach and the whole early warning level of the current river reach, then determining the possibility of various early warning levels under the current real-time data by combining historical data, and taking the possibility as a first constraint quantity;
determining the possibility of early warning in each month according to historical data, and taking the possibility as a second constraint quantity;
respectively establishing daily average rainfall capacity and early warning grade relation curves of the current river reach and other areas, calculating the similarity between the corresponding relation curves of the current river reach and other areas, determining other areas similar to the early warning condition of the current river reach and taking the other areas as reliable areas, calculating the difference value between the historical daily average rainfall capacity corresponding to the reliable areas when the early warning grades are generated and the current river reach rainfall capacity, and taking the difference value as a third constraint quantity;
and constructing a loss function of the neural network according to the first constraint quantity, the second constraint quantity and the third constraint quantity, finishing the training of the neural network according to the constructed loss function, inputting hydrologic data of the current river reach acquired in real time into the trained neural network to obtain a corresponding early warning grade, and finishing the flood peak early warning of the current river reach.
2. The flood peak early warning method based on the internet of things and big data as claimed in claim 1, wherein the determining process of the first constraint quantity is as follows:
calculating the importance degree of the water level data obtained on each water level meter:
Figure DEST_PATH_IMAGE001
wherein the content of the first and second substances,
Figure 671018DEST_PATH_IMAGE002
SW is the water level of the water level meter, SJ is the probability of the category to which the current water level belongs after the water levels of all the water level meters are clustered,
Figure 81271DEST_PATH_IMAGE003
indicates the maximum value among all the water levels,
Figure 601245DEST_PATH_IMAGE004
representing the maximum value of the occurrence probability of various water levels;
determining the deviation degree of the early warning level of the water level corresponding to each water level gauge and the early warning level of the whole water level of the river reach:
Figure 198580DEST_PATH_IMAGE005
Figure 859106DEST_PATH_IMAGE006
wherein the content of the first and second substances,
Figure 858286DEST_PATH_IMAGE007
the deviation degree between the water level early warning level corresponding to the ith water level meter and the whole early warning level of the river reach is shown, T is the total number of the water level early warning levels obtained by the ith water level meter in a set time period,
Figure 80320DEST_PATH_IMAGE008
the number of times that the early warning level of the water level corresponding to the ith water level gauge is inconsistent with the integral early warning level of the river reach of the drainage basin is represented,
Figure 666416DEST_PATH_IMAGE009
indicating the deviation of the early warning level of the water level corresponding to the ith water level meter and the integral early warning level of the river reach,
Figure 100939DEST_PATH_IMAGE010
showing the correlation between the j-th water level early warning level on the ith water level meter and the whole early warning level of the river reach,
Figure 220205DEST_PATH_IMAGE011
indicating the j-th water level early warning level on the ith water level meter,
Figure 347561DEST_PATH_IMAGE012
representing the whole early warning level of the river reach of the river basin;
determining the possibility of various early warning levels under the current real-time data by combining historical data:
Figure 418023DEST_PATH_IMAGE013
Figure 656237DEST_PATH_IMAGE014
wherein, the first and the second end of the pipe are connected with each other,
Figure DEST_PATH_IMAGE015
indicating the likelihood of an early warning level of p,
Figure 797719DEST_PATH_IMAGE016
the distance between the water level corresponding to the estimated water level line and the warning line is obtained by the sum of the water level at the ith water level meter and the rainfall which are acquired in real time,
Figure DEST_PATH_IMAGE017
the distance between the water level corresponding to the early warning level on the ith water level meter and the warning line when the early warning level of the whole river basin in the historical data is p,
Figure 768080DEST_PATH_IMAGE018
the total years of the history data used, Q is the number of water level gauges,
Figure 325838DEST_PATH_IMAGE019
in order to be the age-related weight,
Figure 367743DEST_PATH_IMAGE020
in order to be a threshold value for the confidence of the data,
Figure 196022DEST_PATH_IMAGE021
is the first in the history data
Figure 930760DEST_PATH_IMAGE022
Early warning level of year.
3. The flood peak early warning method based on the internet of things and big data as claimed in claim 2, wherein the second constraint quantity is determined by the following process:
the probability of early warning occurring in each month is:
Figure 713164DEST_PATH_IMAGE023
Figure 558761DEST_PATH_IMAGE024
wherein the content of the first and second substances,
Figure 507125DEST_PATH_IMAGE025
indicating the likelihood of a pre-alarm occurring at month m, Y is the total years of historical data employed,
Figure 645720DEST_PATH_IMAGE026
indicating the number of occurrences of the pre-warning at the mth month of the u-th year,
Figure DEST_PATH_IMAGE027
indicating the total number of occurrences of the pre-alarm in the u-th year,
Figure 351639DEST_PATH_IMAGE019
in order to be the age-related weight,
Figure 735347DEST_PATH_IMAGE020
in order to be a threshold value for the confidence of the data,
Figure 33823DEST_PATH_IMAGE021
is the first in the history data
Figure 110364DEST_PATH_IMAGE022
Early warning level of year.
4. The flood peak early warning method based on the internet of things and big data as claimed in claim 3, wherein the third constraint quantity is determined by the following process:
establishing a relation curve between daily average rainfall and early warning level of the current river reach and other areas, and calculating a dynamic time normalization distance between the corresponding relation curves of the current river reach and other areas:
Figure 631475DEST_PATH_IMAGE028
wherein, the first and the second end of the pipe are connected with each other,
Figure 51830DEST_PATH_IMAGE029
the dynamic time integration distance between the average rainfall-early warning grade curve of the current river reach in the past year and the average rainfall-early warning grade curve of other areas in the past year is shown, R is the total number of element points on the average rainfall-early warning grade curve in the past year,
Figure 974786DEST_PATH_IMAGE030
is the minimum cost between corresponding points of the two curves;
determining the reliability of other areas same as the rainfall condition of the current river reach:
Figure 691070DEST_PATH_IMAGE031
wherein E is the reliability of the rainfall condition of the current river reach identical to that of other areas, Y is the total years of the adopted historical data,
Figure 433898DEST_PATH_IMAGE032
the dynamic time normalization distance between the daily average rainfall-early warning grade curve of the current river reach corresponding to the u-th year in history and the daily average rainfall-early warning grade curve of other areas is calculated;
and taking other corresponding areas when the reliability is greater than the reliability threshold value as reliable areas, and calculating the difference between the rainfall amount corresponding to each early warning level of the reliable areas and the rainfall amount acquired in real time by the current river reach:
Figure DEST_PATH_IMAGE033
wherein the content of the first and second substances,
Figure 332977DEST_PATH_IMAGE034
the difference value between the historical daily average rainfall and the current river reach rainfall collected in real time when the early warning grade of each reliable area is p,
Figure 844861DEST_PATH_IMAGE035
for the number of reliable areas to be used,
Figure 496160DEST_PATH_IMAGE036
is as follows
Figure DEST_PATH_IMAGE037
The reliability of the individual reliable regions is,
Figure 132809DEST_PATH_IMAGE038
is a first
Figure 181053DEST_PATH_IMAGE037
When the early warning level of each reliable area is pThe historical daily average rainfall of the rain,
Figure 547444DEST_PATH_IMAGE039
the rainfall of the current river reach is collected in real time.
5. The flood peak early warning method based on the internet of things and big data as claimed in claim 4, wherein the loss function is:
Figure 871109DEST_PATH_IMAGE040
Figure 87064DEST_PATH_IMAGE041
Figure 888798DEST_PATH_IMAGE042
Figure 109695DEST_PATH_IMAGE043
wherein the content of the first and second substances,
Figure 371305DEST_PATH_IMAGE044
in order to be a function of the loss of the neural network,
Figure 310443DEST_PATH_IMAGE045
is a cross entropy loss function, namely a loss function formed by the early warning level output by the neural network and the early warning level of the label corresponding to the input data,
Figure 915867DEST_PATH_IMAGE046
when the output result of the neural network is the early warning grade p, the whole early warning grade of the river basin obtained by using the historical data for artificial analysis is pThe performance of the composite material is improved,
Figure 755385DEST_PATH_IMAGE047
indicating the possibility of the early warning result output by the neural network occurring in the current month and the possibility of the no early warning result output by the neural network occurring in the current month,
Figure 420853DEST_PATH_IMAGE048
the difference value between the historical daily average rainfall and the current river reach rainfall collected in real time when the early warning grade of each reliable area is p,
Figure 581707DEST_PATH_IMAGE049
early warning the flood peak for the possibility of early warning in the current month,
Figure 752007DEST_PATH_IMAGE050
the times of early warning the flood peak in the current month with the early warning level more than 1,
Figure DEST_PATH_IMAGE051
the total days of the month are pre-warned for the flood peak,
Figure 619600DEST_PATH_IMAGE052
the days that the month has elapsed are warned of the flood peak.
CN202210953183.3A 2022-08-10 2022-08-10 Flood peak early warning method based on Internet of things and big data Active CN115014299B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210953183.3A CN115014299B (en) 2022-08-10 2022-08-10 Flood peak early warning method based on Internet of things and big data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210953183.3A CN115014299B (en) 2022-08-10 2022-08-10 Flood peak early warning method based on Internet of things and big data

Publications (2)

Publication Number Publication Date
CN115014299A true CN115014299A (en) 2022-09-06
CN115014299B CN115014299B (en) 2022-11-18

Family

ID=83066217

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210953183.3A Active CN115014299B (en) 2022-08-10 2022-08-10 Flood peak early warning method based on Internet of things and big data

Country Status (1)

Country Link
CN (1) CN115014299B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115661438A (en) * 2022-10-21 2023-01-31 广州地铁集团有限公司 Outdoor water level monitoring and flood prevention early warning method based on machine vision
CN116046079A (en) * 2023-04-03 2023-05-02 山东省地质调查院(山东省自然资源厅矿产勘查技术指导中心) Geological environment thematic data integrated management system based on position

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016198873A1 (en) * 2015-06-09 2016-12-15 Smith Alexis Hannah Flood risk mapping and warning system and method
JP2019095240A (en) * 2017-11-20 2019-06-20 日本無線株式会社 Water level prediction method, water level prediction program, and water level prediction device
CN110737875A (en) * 2019-09-24 2020-01-31 浙江省水利河口研究院 medium-and-long-term torrential flood disaster early warning and prejudging method
CN111141261A (en) * 2019-12-19 2020-05-12 紫光云技术有限公司 Hydrologic monitoring data acquisition and transmission method
CN111814407A (en) * 2020-07-28 2020-10-23 安徽沃特水务科技有限公司 Flood forecasting method based on big data and deep learning
CN112327698A (en) * 2020-11-05 2021-02-05 叶远 Flood disaster early warning system and method based on Internet of things
CN113379098A (en) * 2021-04-16 2021-09-10 水利部珠江水利委员会技术咨询(广州)有限公司 River interval dynamic flood peak analysis early warning method, system, equipment and medium
CN113780668A (en) * 2021-09-15 2021-12-10 泰华智慧产业集团股份有限公司 Urban ponding waterlogging prediction method and system based on historical data

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016198873A1 (en) * 2015-06-09 2016-12-15 Smith Alexis Hannah Flood risk mapping and warning system and method
JP2019095240A (en) * 2017-11-20 2019-06-20 日本無線株式会社 Water level prediction method, water level prediction program, and water level prediction device
CN110737875A (en) * 2019-09-24 2020-01-31 浙江省水利河口研究院 medium-and-long-term torrential flood disaster early warning and prejudging method
CN111141261A (en) * 2019-12-19 2020-05-12 紫光云技术有限公司 Hydrologic monitoring data acquisition and transmission method
CN111814407A (en) * 2020-07-28 2020-10-23 安徽沃特水务科技有限公司 Flood forecasting method based on big data and deep learning
CN112327698A (en) * 2020-11-05 2021-02-05 叶远 Flood disaster early warning system and method based on Internet of things
CN113379098A (en) * 2021-04-16 2021-09-10 水利部珠江水利委员会技术咨询(广州)有限公司 River interval dynamic flood peak analysis early warning method, system, equipment and medium
CN113780668A (en) * 2021-09-15 2021-12-10 泰华智慧产业集团股份有限公司 Urban ponding waterlogging prediction method and system based on historical data

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
RAMLI ADNAN: "Flood Water Level Modelling and Prediction Using Artificial Neural Network : Case Study of Sungai Batu Pahat in Johor", 《IEEE CONTROL AND SYSTEM GRADUATE RESEARCH COLLOQUIUM》 *
王一峰等: "基于人工神经网络的倒水河洪水预报模型", 《水资源研究》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115661438A (en) * 2022-10-21 2023-01-31 广州地铁集团有限公司 Outdoor water level monitoring and flood prevention early warning method based on machine vision
CN115661438B (en) * 2022-10-21 2024-03-08 广州地铁集团有限公司 Off-site water level monitoring and flood prevention early warning method based on machine vision
CN116046079A (en) * 2023-04-03 2023-05-02 山东省地质调查院(山东省自然资源厅矿产勘查技术指导中心) Geological environment thematic data integrated management system based on position
CN116046079B (en) * 2023-04-03 2023-06-30 山东省地质调查院(山东省自然资源厅矿产勘查技术指导中心) Geological environment thematic data integrated management system based on position

Also Published As

Publication number Publication date
CN115014299B (en) 2022-11-18

Similar Documents

Publication Publication Date Title
CN115014299B (en) Flood peak early warning method based on Internet of things and big data
CN111047099B (en) Regional torrential flood risk prediction method and system
CN112785043B (en) Flood forecasting method based on time sequence attention mechanism
CN104408900B (en) Neutral net flood warning devices and methods therefor based on dynamic optimization
CN106650767B (en) Flood forecasting method based on cluster analysis and real-time correction
CN110298480B (en) Method and system for calculating critical rainfall index of mountain torrent disaster
CN106373070A (en) Four-prevention method for responding to city rainstorm waterlogging
CN112016831A (en) AI intelligent forecast-based urban waterlogging landing area identification method
CN113095694B (en) Rainfall sand transportation model construction method suitable for multiple landform type areas
CN117113236B (en) Smart city monitoring system and data processing method
CN106228268A (en) A kind of Forecasting Methodology of expressway travel time based on charge data
Priatna et al. Precipitation prediction using recurrent neural networks and long short-term memory
Abraham et al. Comparative Analysis of Various Machine Learning Techniques for Flood Prediction
CN116110210B (en) Data-driven landslide hazard auxiliary decision-making method in complex environment
CN112818668A (en) Meteorological disaster data semantic recognition analysis method and system
CN114638441B (en) Ocean current monitoring and early warning system based on satellite remote sensing image
CN109978263B (en) Garden water system water level early warning method
CN117033935A (en) Prediction method of rainfall characteristic under statistics and monitoring based on Bayesian fusion
CN109508810A (en) A kind of system based on realization monthly average hydrology volume forecasting
CN209785224U (en) inside water level early warning system in gardens
CN114021789A (en) Rainfall prediction method based on characteristic engineering and TabNet network
CN113011657A (en) Method and device for predicting typhoon water level, electronic equipment and storage medium
CN112380778A (en) Weather drought forecasting method based on sea temperature
Zhou et al. Real‐time prediction and ponding process early warning method at urban flood points based on different deep learning methods
Joo et al. Analysis of the optimal window size of hampel filter for calibration of real-time water level in agricultural reservoirs

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
PE01 Entry into force of the registration of the contract for pledge of patent right

Denomination of invention: Flood peak warning method based on the Internet of Things and Big data

Effective date of registration: 20230704

Granted publication date: 20221118

Pledgee: Xi'an innovation financing Company limited by guarantee

Pledgor: Mountain Technology Co.,Ltd.

Registration number: Y2023610000517

PE01 Entry into force of the registration of the contract for pledge of patent right